OBJECTIVES: This study aims to guide researchers on the best way to score the Center for Epidemiologic Studies – Depression scale (CES-D) by comparing competing models in the literature. METHODS: Radloff (19777), the original CES-D author, first provided scoring for four uncorrelated factors of depression: negative affect (NA), positive affect (PA), interpersonal (I), and somatic (S).  Sheehan et al. (1995) validated four correlated factors in the CES-D.  However, Radloff and Sheehan results may have differed because Radloff ascribed items of failure and fearful to NA whereas Sheehan assigned a value of I.  Following Sheehan’s scoring, Cole et al. (2004) presumed a hierarchical factor of depressed mood that included all four previously identified factors and posited a 10-item short-form of the CES-D.  These four models were compared using structural equation modeling for parametric models with Bollen-Stine bootstraps to control for multivariate nonnormality in 225 community-residing subjects, structural validity of the models were compared. RESULTS: Fit statistics for the four models were: Radloff comparative fit index (CFI) = .790 & root mean square error of approximation (RMEA) = .011; Sheehan CFI = .926 & RMSEA = .053, Cole hierarchical CFI = .927 & RMSEA = .052; Cole 10-item CFI = .979 & RMSEA = .041 CONCLUSIONS: The uncorrelated Radloff model was the poorest fit the day.  Both Sheehan and Cole 20-item models were decently fit to the data and nearly identical to each other in fit.  Finally, the 10-item Cole short-form of the CES-D was the best fit model to the data.  Given the brevity of this form and strong fit with the theoretical structure postulated by Radloff, researchers may want to consider this form of the CES-D for research on depressed mood.